%% Solve for the steady state of my model
% with search frictions, menu costs
% and idiosyncratic productivity
% fixed menu cost
% approximating the value of adjustment, value of not adjustment, and the
% expected value function 

%addpath ~/compecon/CEtools

soblen = 100;

%% Set Options
% All
options.Nbell       = 100;        % Number of Bellman iterations before Newton.
options.Nnewt       = 15;       % Maximum number of Newton steps
options.tolc        = 1e-12;     % Tol on value functions
options.T_irf       = 50;       % Number of periods for IRFs
options.T           = 300;     % Number of periods for MIT shock
options.Terg        = 500;      % Number of periods to burn
options.hpisteps    = 100;

% Stationary
options.L = [];
options.itermaxL    = 5000;     % Max number of iterations to find L
options.tolL        = 1e-12;    % Tol for L
options.tolK        = 1e-5;     % Tol for equilibrium K
options.itermaxK    = 100;      % Max iterations for bisection
options.cresult = [];

options.tolD        = 1e-08;
options.itermaxp    = 25;

options.damp = 0;
options.damp_mit = .995;
options.damp_DC  = .9;

optset('bisect', 'tol', 1e-12)

%% Set globals and parameters
% % try vavra numbers
glob.n          = [50,30];   % Number of nodes for b,a
glob.nf         = [50,30];

glob.spliorder  = [1,1];    % Order of splines (Envelope Condition Method seems only robust when quadratic in k)
glob.Na         = 30;           % Number of nodes for quadrature
glob.curv       = .1;            % Amount of curvature for p grid

glob.Nsignal    = 1e6;

glob.minb = 0;
glob.maxb = 15;


%% Model parameters

% preferences
param.beta      = 0.96;               

%  rho_s    phi_L  sigma epsilon   ce_mu     d_E
% 0.60338 0.076289 2.9199   2.835 -2.3445 0.59226

% idiosyncratic shock - set 'em and forget 'em
param.mu_s           = 0; % Cooper Haltiwanger (2002)
% note - berger just chooses rho - .81 and sigma = .38 (annual)

%%%%%%%%%%%%% CALIBRATE HERE %%%%%%%%%%%%%
param.sigma_s         = 0.18; % concentration & variance
param.rho_s           = 0.79;  % Concentration & variance
 
% labor adjustment cost
param.phi_L           = 0.07; % as a start, no adjustment costs
 
% Demand parameters
param.sigma           = 20;              % FIX THIS NUMBER, let go of average markup
param.epsilon         = .6*param.sigma;     % epsilon/sigma is superelasticity, 0 is CES 
 
% fixed cost of production
param.mu_f     = -1e6; 2.15;
param.sigma_f  = 1.65;
 
% dist of entrant productivity signals
param.xi     = .95;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

param.omega           = 1;                  
         
param.A = 1;

% entry and exit
param.gamma = .015;     % exogenous exit rate

% sunk cost of entry 
param.ce       = exp(param.mu_f + param.sigma_f^2/2);
param.ce_noise = 1e-3;

param.M        = 1;


param.nu      = .5; % inverse Frisch elasticity
                     % same as in clementi
                     % can calibrate aggregate shock size to hit moments

% free labor adjustment 
param.delta    = 0.1;

glob.mink       = 0;
glob.maxk       = 1e2;


%% calibrate
options.itermax_DC = 100;

P = sobolset(6);
sobolstart = (soblen-1)*sobnum + 1;
sobolend = sobolstart + soblen;

P = P(sobolstart:sobolend,:);

%    rho,   sigma_s, sigma, sigma/epsilon,  xi      phi_L
lb = [.75   .13       15      .45           .5       .03];
ub = [.83   .23       25      .75            1.5     .15];

loc = lb;
scale = ub - lb;

C = [];
M = [];

for p = 1:soblen
    calib = P(p,:);
    calib = loc + scale.*calib
    [mom] = compute_moments_for_calibration(calib, param, glob, options);

    if flag
       continue 
    end
    
    dlmwrite(['results/calib', num2str(sobnum), '.csv'],calib,'delimiter', ',','-append');
    
    f = fields(mom);
    mom_out = nan(length(f),1);
    for i = 1:length(f)
        mom_out(i) = mom.(f{i});
    end
    
    dlmwrite(['results/moments', num2str(sobnum), '.csv'],mom_out','delimiter',',','-append');

end

